-
Notifications
You must be signed in to change notification settings - Fork 496
Expand file tree
/
Copy pathGPUCommonAlgorithmThrust.h
More file actions
122 lines (104 loc) · 3.59 KB
/
GPUCommonAlgorithmThrust.h
File metadata and controls
122 lines (104 loc) · 3.59 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
// Copyright 2019-2020 CERN and copyright holders of ALICE O2.
// See https://alice-o2.web.cern.ch/copyright for details of the copyright holders.
// All rights not expressly granted are reserved.
//
// This software is distributed under the terms of the GNU General Public
// License v3 (GPL Version 3), copied verbatim in the file "COPYING".
//
// In applying this license CERN does not waive the privileges and immunities
// granted to it by virtue of its status as an Intergovernmental Organization
// or submit itself to any jurisdiction.
/// \file GPUCommonAlgorithmThrust.h
/// \author Michael Lettrich
#ifndef GPUCOMMONALGORITHMTHRUST_H
#define GPUCOMMONALGORITHMTHRUST_H
#ifndef GPUCA_GPUCODE_COMPILEKERNELS
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wshadow"
#include <thrust/sort.h>
#include <thrust/execution_policy.h>
#include <thrust/device_ptr.h>
#pragma GCC diagnostic pop
#include "GPUCommonDef.h"
#include "GPUCommonHelpers.h"
#ifndef __HIPCC__ // CUDA
#include <cub/cub.cuh>
#else // HIP
#include <hipcub/hipcub.hpp>
#endif
#endif // GPUCA_GPUCODE_COMPILEKERNELS
#ifndef __HIPCC__ // CUDA
#define GPUCA_THRUST_NAMESPACE thrust::cuda
#define GPUCA_CUB_NAMESPACE cub
#else // HIP
#define GPUCA_THRUST_NAMESPACE thrust::hip
#define GPUCA_CUB_NAMESPACE hipcub
#endif
namespace o2::gpu
{
// - Our quicksort and bubble sort implementations are faster
/*
template <class T>
GPUdi() void GPUCommonAlgorithm::sort(T* begin, T* end)
{
thrust::device_ptr<T> thrustBegin(begin);
thrust::device_ptr<T> thrustEnd(end);
thrust::sort(thrust::seq, thrustBegin, thrustEnd);
}
template <class T, class S>
GPUdi() void GPUCommonAlgorithm::sort(T* begin, T* end, const S& comp)
{
thrust::device_ptr<T> thrustBegin(begin);
thrust::device_ptr<T> thrustEnd(end);
thrust::sort(thrust::seq, thrustBegin, thrustEnd, comp);
}
template <class T>
GPUdi() void GPUCommonAlgorithm::sortInBlock(T* begin, T* end) // TODO: Try cub::BlockMergeSort
{
if (get_local_id(0) == 0) {
sortDeviceDynamic(begin, end);
}
}
template <class T, class S>
GPUdi() void GPUCommonAlgorithm::sortInBlock(T* begin, T* end, const S& comp)
{
if (get_local_id(0) == 0) {
sortDeviceDynamic(begin, end, comp);
}
}
*/
template <class T>
GPUdi() void GPUCommonAlgorithm::sortDeviceDynamic(T* begin, T* end)
{
thrust::device_ptr<T> thrustBegin(begin);
thrust::device_ptr<T> thrustEnd(end);
thrust::sort(GPUCA_THRUST_NAMESPACE::par, thrustBegin, thrustEnd);
}
template <class T, class S>
GPUdi() void GPUCommonAlgorithm::sortDeviceDynamic(T* begin, T* end, const S& comp)
{
thrust::device_ptr<T> thrustBegin(begin);
thrust::device_ptr<T> thrustEnd(end);
thrust::sort(GPUCA_THRUST_NAMESPACE::par, thrustBegin, thrustEnd, comp);
}
#ifndef GPUCA_GPUCODE_COMPILEKERNELS
template <class T, class S>
GPUhi() void GPUCommonAlgorithm::sortOnDevice(auto* rec, int32_t stream, T* begin, size_t N, const S& comp)
{
thrust::device_ptr<T> p(begin);
#if 0 // Use Thrust
auto alloc = rec->getThrustVolatileDeviceAllocator();
thrust::sort(GPUCA_THRUST_NAMESPACE::par(alloc).on(rec->mInternals->Streams[stream]), p, p + N, comp);
#else // Use CUB
size_t tempSize = 0;
void* tempMem = nullptr;
GPUChkErrS(GPUCA_CUB_NAMESPACE::DeviceMergeSort::SortKeys(tempMem, tempSize, begin, N, comp, rec->mInternals->Streams[stream]));
tempMem = rec->AllocateVolatileDeviceMemory(tempSize);
GPUChkErrS(GPUCA_CUB_NAMESPACE::DeviceMergeSort::SortKeys(tempMem, tempSize, begin, N, comp, rec->mInternals->Streams[stream]));
#endif
}
#endif // #ifndef GPUCA_GPUCODE_COMPILEKERNELS
} // namespace o2::gpu
#undef GPUCA_THRUST_NAMESPACE
#undef GPUCA_CUB_NAMESPACE
#endif